Thunderstorm (1959) by Blair Rowlands Hughes-Stanton.
• “To create a novel or a painting, an artist makes choices that are fundamentally alien to artificial intelligence,” says SF writer Ted Chiang. A New Yorker essay which has received a fair amount of attention over the past week, with good reason. As someone who found his name on the list of artists whose work was allegedly being fed into Midjourney, I suppose I have a vested interest in the arguments. (Good luck to any machine trying to imitate my “style”. I don’t have one.) Too much of the discussion, however, has been very poor which is why this is the first time I’ve linked to such a piece here.
• “After going their own way for much of the 20th century, mathematicians are increasingly turning to the laws and patterns of the natural world for inspiration. Fields stuck for decades are being unstuck. And even philosophers have started to delve into the mystery of why physics is proving ‘unreasonably effective’ in mathematics, as one has boldly declared.” Ananyo Bhattacharya on why physics is good at creating new mathematics. Having recently finished reading Cormac McCarthy’s final novel, Stella Maris, this was all very timely.
• “…our films obey musical laws. Of course, you can never tell people how they should watch a film. But the musical element provides a narrative of its own.” Thus the Quay Brothers, in the news again with their forthcoming feature film, Sanatorium Under the Sign of the Hourglass. The quote is from a recent interview with Xan Brooks. Meanwhile, Alex Dudok de Wit posted another interview from 2019, originally published in French, now made available in English for the first time.
• At Wormwoodiana: Mark Valentine announces a new book of his essays, The Thunderstorm Collectors.
• At Dennis Cooper’s: 28 books that either faked ingesting LSD or did.
• At Public Domain Review: Antiquities of Mexico (1831–48).
• At Print mag: Kelly Thorn’s Tarot of Oxalia.
• USC Optical Sound Effects Library
• Strange Thunder (1987) by Harold Budd | Sweet Thunder (1991) by Yello | Studies For Thunder (2004) by Robert Henke
“As someone who found his name on the list of artists whose work was allegedly being fed into Midjourney, I suppose I have a vested interest in the arguments. (Good luck to any machine trying to imitate my “style”. I don’t have one.)”
If it was me, I’d be knocking on wood, crossing fingers, setting up a shrine, whatever.
Everything being predicted beyond complex calculations or, arts-wise, involves ripping off the most famous creators is some times off.
So yeah, AI will be able to do art in the style of John Coulthart but no time soon. (Caveat: all this “no time soon” stuff actually happens sometimes between when those with vested interests in hyping AI and, well, no time soon. So in this case, I’d expect AI to do at least good enough image making in the style of Coulthart but, well, not at any time that it would be a problem.
OTOH, maybe I’m misinterpreting John in the quote. Maybe he accepts AI being able to render images that looks like his art except that they’re really crappy.
Which gets to the uber-problem to AI now: by and by, it can’t separate BS from the real thing. So, sure, it can render a Coulthartesque image but nothing as good as the real thing.
Some day it’ll be scary good. But not so soon as to stress over it.
I hope.
I’m not blasé about any of this, just sceptical about many of the claims and the suggestion that quality/ability will continue to improve in an exponential manner. This last seems unlikely because none of these systems know what anything is, as in really know, the way even small children and the smarter animals do. That’s why we’ve had so many AI pictures getting fundamental aspects of human anatomy wrong. They copy a hand but they don’t know what “a hand” is. Now multiply “hand” into everything else in the world and the history of humanity…
As for my own art, you could probably generate variations of a single picture or design right now but I’ve never worked in a single recognisable style that can be easily co-opted. That’s what I mean why I say I don’t have a style. Everything in the link below is based around the same subject, all the art is my own and yet all the pieces are very different to each other:
https://www.johncoulthart.com/pantechnicon/cthulhucalendar.html
I’ve often thought it was a drawback being so versatile but this is one area where you have a slight advantage.
There was a game played in the 18th century, I think, that involved creating a story plot using cards that could be rearranged in a finite but very large number of possibilities. That’s if I remember it right. I think similar things have been created or been used themselves as plot devices. And it’s how, I suppose, the tarot and i ching compose.
AI works – as in functionally devises, it doesn’t really ‘work’ at all – on the same principle. It has no ‘intelligence’, and as John said, knows nothing. It’s an enormous reference arranger, nothing more. Why it messes stuff up when it recomposes from its atomised pool of referents I presume is down to the design (ie, not the user’s) ‘instructions’ which I think are still secret. The best the fans can hope for is that these will refine and iron out glitches in its effluent.
It will always be a mere simulacrum because it can never have a qualitative basis of experience, and therefore can never know or understand anything. No matter how perfected, like the mechanical nightingale. A basic ‘hello, it’s a nice day’ from a human has worlds within it, an AI ‘Shakespeare’ never can. Same with pictures.
Thinking it has any more is the crudest superstition, magical thinking/demonic inhabiting/synchronicity/sympathies, etc.
No doubt many will plump for that angle, if they haven’t already.